1. 云计算入门 Introduction to Cloud Computing GESC1001
Philippe Fournier-Viger
Professor
School of Humanities and Social Sciences
大家好！ Yun2 ji4 suan4 ru4men2
Fall 2017

2. About me Born in Canada Ph.D. in Computer Science (Canada) (2010)
Post-doctoral researcher at 成功大学 (2010)
Professor at U. of Moncton, Canada ( ),
Professor (教授) at HITSZ (哈尔滨工业大学 (深圳) ( …), and Youth 1000 talent (青年千人)
My research: data mining (数据挖掘 \ 大数据).
I speak English, French, and basic Chinese.

3. How to contact me? We can discuss immediately after lectures
Office: G421 office phone:
Wechat: philfv
I check my s once a day

4. Teaching assistants
The two teaching assistants: Zhang Yimin (张毅民), Li Zhitian (李之天)
will:
grade homeworks,
check the attendance,
help with other issues related to the course.
QQ group:

5. About this course Evaluation: 16 hours (8 lectures)
1 credit, general course (本科专业课程, 通识课程 ）
No requirements
Evaluation:
Attendance (10 %)
Course work (30%: two assignments - 作业)
Final exam (60%)
Liangge zuoye

6. Main objective
Understand what is cloud computing ( 云计算), how it works, how it is used, and when it should be used.
Examples:
Baidu cloud (百度云)
Alibaba Cloud (阿里云)
Scientific applications
Web search engines (搜索引擎)
云计算 Yun2ji4suan4 = cloud computing

7. Specific objectives Understand what is cloud computing (云计算)
Understand challenges and solutions for parallel computing (并行计算) and distributed computing (分布式计算 ).
Understand the infrastructure (基础设施) required for cloud computing 

8. Specific objectives
Understand how applications are developed for cloud computing
Understand other aspects:
Security
Virtualization (虚拟化)
How data is stored in the cloud
Current trends and technologies

9. Course schedule Week 1 Introduction and overview Week 2
Distributed and parallel systems
Week 3
Cloud infrastructure
Week 4
Cloud application paradigm
Week 5
National holiday
Week 6
Cloud application paradigm (2)
Week7
Cloud virtualization and resource management
Week 8
Cloud computing storage systems
Week 9
Cloud computing security
Week 10
Final exam

10. Textbook I will provide detailed lecture slides (PPTs)
I will use this book as main reference to prepare the course:
D. C. Marinescu. Cloud Computing Theory and Practice, Morgan Kaufmann, 2013.
It is not necessary to buy the book! (没有必要买这本书)

11. The course website
PPTs, assignments (两个作业), grades (分数), and the course schedule are provided on the course website:

13. 课程内容

17. Rules Students must attend all lectures. Students must arrive on time.
Homework must be submitted before the deadline. No late homework will be accepted.

18. Rules Plagiarism (抄袭 ) will not be tolerated.
Phones should be turned off during lectures
One should not talk or eat in the classroom, or disturb other classmates.

19. Lectures Each lecture starts at 18:30 Each lecture ends at 20:05
We will take a 5 minute break around 19:20.

20. Chapter 1 - Introduction
pdf 20-38

21. Introduction A computer can perform simple tasks: A fast computer can:
e.g. play music, video games, video editing, simulations, writing documents
A fast computer can:
perform more complex tasks
perform them in less time.
Every year, more “powerful” computers are sold.

22. How powerful a computer is?
It can be measured in terms of:
speed: the number of instructions (指令) that the computer processes per second,
data processing speed (数据处理速度): the amount of data that it can process per second,
storage space (存储空间), the amount of data that it can store,
network speed (网络速度), its ability to receive and send data to other computers …

23. Supercomputers (超级计算机)
Fast, powerful computers, expensive (昂贵的)
Speed: can be more than 1 trillion instructions per second,
Weight: can be more than 100 tons
Data: may store more than 20,000 times more data than a desktop computer (台式电脑)
A super-computer in Czech Republic
捷克共和国
Radio.cz
Used for: medicine simulation, weather forecasting, nuclear research,….

24. A problem! I have a computer that is slow!
I would like to do some very complex tasks (e.g. weather prediction 天气预报).
But, I don’t want to buy a new computer.
Idea: could I rent one or more computers to do the task (能租电脑)?

25. A solution
I would like to be able to rent computers when I need more computing power （ 计算能力 ） or storage space （ 存储空间).
I want to pay only for what I use.
This is the main idea of cloud computing (云 计算)
Similar to electricity (电) or water (水):
I can buy more electricity when I need it
I don’t need to know how the electricity is made, transported, or where it come from.
I just pay for the electricity that I use.

26. Cloud computing (云计算)
Using a cloud (a set of distant computers) to perform computing tasks.
CLOUD
I will provide a more accurate definition later.

27. Cloud (云)
A set of distant (遥远的) computers that provide computing or storage services to users.
Mi Cloud (小米云),
Baidu cloud (百度云),
Alibaba cloud (阿里云) …
CLOUD
I will provide a more accurate definition later.

28. 小米 Cloud provider (云提供商 )
An organization that provides or sell the right to use its cloud.
Xiaomi (小米)
Alibaba (阿里巴巴) … 小米
CLOUD

29. A cloud must be « elastic »
Characteristics of a cloud
A cloud must be « elastic »
Elastic computing (弹性计算) The ability to dynamically acquire computing resources to offer variable workload (变负荷).
CLOUD

30. A cloud must be « elastic »
Elastic computing (弹性计算) The ability to dynamically acquire computing resources to offer variable workload (变负荷).
CLOUD
Customer (顾客)

31. A cloud must be « elastic »
Elastic computing (弹性计算) The ability to dynamically acquire computing resources to offer variable workload (变负荷).
CLOUD
Customers (顾客)

32. A cloud must be « elastic »
Elastic computing (弹性计算)
The ability to dynamically acquire computing resources to offer variable workload (变负荷).
CLOUD
Customers (顾客)

33. A cloud must be « elastic »
To support elastic services, a cloud provider (云提供商) must maintain a massive infrastructure (基础设施).
Resource usage is generally metered (计量) (amount of data, CPU time, storage space used…).
Users are charged ($) for the resources that they use (in general).

34. A cloud must be maintained
Maintenance (维护):
Maintenance and security is the responsibility of the cloud provider (云提供商).
Users
do not need to install, maintain and update computers and software to run applications.
do not need to buy equipment.
Security
expert
Technician
Equipment can be expensive!

35. Cost of running a cloud Cloud providers (云提供商):
operate efficiently due to specialization (专业化) and centralization (集中).
are very effective due to resource multiplexing (资 源复用) (sharing/reusing resources for several users)
CLOUD
9:00 AM
Red = computers used by company 1
Blue = computers used by company 2

36. Cost of running a cloud Cloud providers (云提供商):
operate efficiently due to specialization (专业化) and centralization (集中).
are very effective due to resource multiplexing (资 源复用) (sharing/reusing resources for several users)
CLOUD
9:15 AM
Red = computers used by company 1
Blue = computers used by company 2

37. Cost of running a cloud Cloud providers (云提供商):
operate efficiently due to specialization (专业化) and centralization (集中).
are very effective due to resource multiplexing (资 源复用) (sharing/reusing resources for several users)
CLOUD
9:25 AM
Red = computers used by company 1
Blue = computers used by company 2

38. The data is distributed in the cloud
Distributed data storage (分布式数据存储):
Data is stored on multiple computers, often close to the places where it is used.
This increases reliability (可靠性) and security.
This reduces communication costs.
CLOUD
Beijing
Shenzhen
Chengdu
Shenzhen
Beijing
Xian
Qingdao
Dalian
Xian

39. The data is replicated in the cloud
Characteristics of a cloud
The data is replicated in the cloud
Data replication (数据复制):
Data is generally replicated (复制 ) on multiple computers to reduce the risk of losing data, and increase the access speed in different geographical areas (地理区域).
CLOUD
Beijing
Shenzhen
Chengdu
Shenzhen
Beijing
Dalian
Xian
Qingdao
Xian

40. The data is replicated in the cloud
Characteristics of a cloud
The data is replicated in the cloud
Data replication (数据复制):
Data is generally replicated (复制 ) on multiple computers to reduce the risk of losing data, and increase the access speed in different geographical areas (地理区域).
CLOUD
Beijing
Shenzhen
Chengdu
Shenzhen
Beijing
Dalian
Xian
Qingdao
Xian

41. Cloud computing is good?
Yes, many benefits.
However,
Users store private data and software on the cloud.
Users do not have full control of the cloud.
Cloud computing has implications (启示) that are:
Economical
Social
Ethical (伦理的)
Legal

42. Applications of cloud computing
Scientific and engineering applications
Finance,
Gaming,
Weather prediction,
Data mining,
and other data intensive activities …

43. MI Cloud (小米云) and similar services
Online service for storing content such as music, videos, and personal data.
Advantages:
Data is accessible from anywhere,
Data is accessible from multiple devices (phone, laptop…)
Easy to share the data
But safe? secure? Still own the data?

44. Cloud for applications
Services such as iCloud or Mi Cloud are used to store data in the cloud.
To reduce communication costs (通信费 用), data could also be processed in the cloud.
Thus, applications (软件程序) are also often executed in the cloud.

45. Data centers (数据中心)
To benefit from an economy of scale (规模经 济), computers are often stored, maintained and managed in a same location (called data centers - 数据中心).
A data center may contain hundreds of computers.

46. How big are data centers?
Large (50,000 computers) vs medium (1,000 computers) data centers
Resource consumption (including energy): 5 times less
Networking cost and storage cost: > 5 times less
Requires less network administrators per number of computers.

47. How many data centers? In 2006, 6,000 data centers in the US
They consumed 61 billion KWh (千瓦时)
This is 1.5 % of the energy consumption of the US
Data centers should be located in cities where electricity is cheap.

48. The World-Wide Web (万维网)
The WWW is some form of cloud computing?
Not really!
It is a large-scale distributed system with data stored on multiple computers and offering various services to users.
Users can find, share and combine information

49. Grid computing (网格计算) The Web is a form of “grid computing”
Grid: an heterogeneous network(异机种网络), geographically distributed, administrated by different organizations.
GRID:
Organization 2
Organization 3
Organization 1

50. Drawbacks (缺点) of grid computing
Latency (延迟) in communication is higher than for a supercomputer.
Transferring data is slower than in a supercomputer.
An heterogeneous network (异机种网络 - grid) makes it difficult
to run applications.
manage the system, perform scheduling (计划), optimize resource allocation (资源分配), fault tolerance (容错), and perform load balancing (负载平衡).
Grids are useful mainly in science and engineering. They do not address the requirements of governments and the industry.

51. Cloud computing Cloud: a set of computers that is homogeneous.
Same type of computers,
Same security,
Same resource management,
Same cost…
Targets enterprise needs
CLOUD

52. What kind of data in a cloud?
Any type (text, audio, etc.)
Data should be:
Treated according to its type
Retrieved from the closest location, or best location in terms of latency, or download time.
Needs services such as:
right management (权利管理),
control of unsuitable content

53. Cloud applications (云应用)
Software programs running in the cloud are called cloud applications (云应用).
Most applications are data intensive and network intensive
The user does not need a powerful computer to use the cloud (e.g. mobile phone)
CLOUD
User

54. Cloud applications (云应用)
The user runs an application to access the services provided by the cloud.
The cloud infrastructure may provide services for coordinating tasks of different users.
User
CLOUD
User
User

55. Challenges of cloud computing
Multiplexing (资源复用): sharing/reusing resources between users typically requires more resources.
The management of many resources is a challenge. Need resource management strategies.
Quality-of-service (服务质量 ) may be hard to achieve in a distributed environment.

56. Challenges of cloud computing
Sharing data is easy in a network. However, it poses security and privacy issues.
Some mechanisms are needed to support access control (访问控制) (who can access what?) and for recording logs (访 问日志) of accesses/changes made by users.

57. What has triggered the cloud revolution?
Many technological advances:
Multi-core processors on a same chip (faster communication between cores using cache)
Novel storage technology (SSD…)
Faster RAM memory, lower cost
Advances in software system, distributed/parallel computing, programming languages…

58. Peer-to-peer systems (对等系统)
Another popular way of organizing computers in a network is peer-to-peer systems (P2P).
e.g. BT (BitTorrent), Kazaa, Napster…
P2P are generally:
self-organized (自组织的) (no central management process),
Decentralized (分散式网络)

59. Peer-to-peer systems (对等系统)
Example: BT (Bit Torrent) Each user may share files with other users.
Each user manages his computer. There is no central computer to manage the P2P network. Is it cloud computing? 

60. P2P is some form of cloud computing?
No!
a cloud is administered by a single organization.
a cloud is homogeneous
In P2P,
there is no central organization.
Each computer is viewed as having the same responsibilities and is managed by its owner.
It is heterogeneous

61. Advantages of P2P The infrastructure is provided by the users.
e.g. the BT network exists because many computers are running the BT software.
Highly decentralized (分散式网络)
Scalable (可扩展的),
Resilient to faults and attacks
Each node (computers) do not consume a lot of bandwidth
May protect the identity and the data of users

62. Disadvantages of P2P P2P systems are hard to manage effectively and
hard to secure (since everything is distributed).
Protect the privacy of users.
P2P networks are often used to carry out illegal activities.

63. Types of clouds Public cloud (公共云) Private cloud (私有云 )
the cloud is available to everyone.
users may need to pay the cloud provider to use the cloud
Private cloud (私有云 )
used by only one organization (e.g. university, government),
may be operated inside the organization
it requires an investment in infrastructures.
usually more secure than using a public cloud.

64. Types of clouds Community cloud (社区云 ): Hybrid cloud:
infrastructure shared by several organizations (a community).
operated by the organizations or a third party
Hybrid cloud:
two or more clouds combined using standardized or proprietary technology

65. Drawbacks (缺点) of the cloud
Availability of service: the cloud provider should be able to guarantee that the cloud will be available at any time
It is often difficult to move from a cloud to another cloud.
Data confidentiality.
Transferring data to the cloud may be slow.
Performance may not be stable.
The user does not own the infrastructures.
Problems in the cloud…

66. Why cloud computing is popular?
Managing a cloud is not so difficult because a cloud is homogeneous (in terms of hardware and software) and administered by one organization.
The cloud is “elastic” : the user can ask for more resources when needed.
Pay for what you use. No need for an up-front investment (前期投资)

67. Three ways of using the cloud (three delivery models）

68. 1st: Software as a service
The user uses applications offered by the cloud.
The user uses an application on his computer to access the cloud (e.g. a Web browser to access 163 ).
The user does not manage the cloud infrastructure (e.g. network, server, operating system, storage).
The user can only manage some application settings. (e.g. how many s are shown per page).

69. 1st: Software as a service
Other examples:
website
Workflow management,
Collaborative applications,
Customer relationship management (CRM),
Financial management…
Social networks, blogs, wiki…

70. 1st: Software as a service
The data is stored in the cloud.
This delivery model is useful when:
many people/organizations use the same product (e.g. 电子邮件)
There are some peaks in the usage of the application (e.g. payroll 工资表).
Some users want to use mobile devices
There is a short term need.

71. 2nd : Platform as a service
The user
can run his own applications in the cloud created or buy applications.
does not control the cloud infrastructure (server, network, operating system, storage…).
has control of:
applications
may control the application hosting environment (session management, content management, sandboxes…)

72. 2nd : Platform as a service
Not useful if:
a specific programming language must be used,
a specific software/hardware configuration is required to achieve optimal performance.
Example:
Useful for software developers that develop applications and then run them in the cloud.

73. 3rd : Infrastructure as a service
The user can run any software (operating system and applications).
The consumer does not control the could infrastructure.
The consumer controls the operating system, software, how the data is stored, and may have limited control over some networking aspects (e.g. firewall).

74. 3rd : Infrastructure as a service
e.g. Alibaba Cloud
Resources are distributed.
Elasticity: dynamic scaling
Pay-as-you-go, variable pricing models
Hardware is shared by multiple users
Useful when:
demand is volatile （挥发性需求）,
a new business needs computing resources,
an organization is quickly expanding.

75. Some vocabulary
Cloud provider: makes services available to consumers. (e.g. Alibaba)
User / service consumer: uses the services of the cloud provider.
Carrier: the intermediary that provides the connectivity between provider and consumer. (e.g. China Telecom)
Cloud provider
consumer
Carrier (China Telecom) 75

76. Conclusion Before next week, join the QQ group: 606909438
make sure that you can access the website

77. References
Chaptre 1. D. C. Marinescu. Cloud Computing Theory and Practice, Morgan Kaufmann, 2013.